A
NATOMICAL AND
F
UNCTIONAL
C
HARACTERIZATION OF THE
M
OUSE
I
NSULAR
C
ORTEX
Daniel Gehrlach
Dissertation
der Fakultät für Biologie
der Ludwig-Maximilians-Universität München
Erstgutachter: Prof. Dr. Rüdiger Klein Zweitgutachter: Prof. Dr. Laura Busse Tag der Einreichung: 16.01.2020
Daniel Gehrlach 16.01.2020
X
A
BSTRACT
The insular cortex (IC) processes sensory information from within and outside the body and has been implicated in emotion regulation and homeostasis. To date, there is no comprehensive connectivity map of the mouse IC. Therefore, in the first part of this thesis, I created a whole-brain connectivity map of long range connections of the mouse IC. The IC was subdivided into three equally long parts (anterior, medial and posterior IC). Then, I performed cell-type specific monosynaptic rabies tracings to quantify afferent connections of excitatory and inhibitory IC neurons, while adeno-associated viral tracings allowed the detection of excitatory efferent axons. My analyses revealed that all parts of the insula are highly interconnected with multiple cortical and subcortical brain regions, which implies a complex integration of multi-sensory and emotional information in each insular subdivision. The anterior IC displays a distinctive connectivity pattern compared to the medial and posterior IC. While the connectivity of the mIC and pIC suggests a primary role in visceral and sensory integration, the aIC seems to play a central role in manipulating goal-directed motor behavior. These results provide an anatomical framework to guide the design of mechanistic investigations as well as working models of insular cortex function.In the second part of this thesis, I combined insights from optogenetic behavior experiments, fiber photometry-based calcium imaging and the above-mentioned anatomical data to reveal a role of the posterior insula in processing aversive sensory stimuli and bodily states. By performing projection-specific optogenetics, I functionally characterized an insula-to-central amygdala pathway and could show that it mediates anxiety-related behaviors, while an independent nucleus accumbens-projecting pathway regulates feeding upon changes in bodily state.
In summary, the data presented in this thesis support a model in which the insular cortex detects aversive internal states and subsequently manipulates behavior, providing a mechanistic framework for IC function. These findings might aid in understanding how alterations in insula circuitry contribute to neuropsychiatric conditions such as obesity or anorexia, addiction and depression.
A
CKNOWLEDGEMENTS
First, I want to thank Nadine Gogolla for giving me the opportunity and trust to help starting up her lab and performing my doctoral studies under her supervision. She was a great mentor that heavily influenced me and my scientific thinking during the last 5 years. On top of that, it was a humbling experience to be one of the first PhD students in a junior group.I would also like to acknowledge my thesis advisory committee members Ilona Grunwald-Kadow and Bastian Hengerer for their valuable input.
I´m deeply thankful for the intense collaboration and teamgeist our lab developed, especially with Alex, Nate, and our (Master-) students Arthur, Daniela, Michaela and Caro. I am also grateful for the support of Marion Ponserre, who shared valuable information and reagents for rabies tracings. Next, I want to thank Tom Gaitanos for helping me with the analysis of the anatomy study. Further, I am thankful for Alja’s and Tom’s helping hands during the revisions for my first paper. Most importantly, I want to thank my parents Cristina and Helmut Gehrlach as well as my wife Ellen for always supporting and believing in me. I would be nothing without them and they deserve all the credit. Thank you!
C
ONTENTS
1 INTRODUCTION ... 1
1.1ANATOMY OF THE RODENT INSULAR CORTEX ... 1
1.1.1 Cyto- and Chemoarchitecture of the Insular Cortex ... 2
1.1.2 Connectivity of the Mouse Insular Cortex ... 2
1.2FUNCTIONS OF THE INSULAR CORTEX ... 3
1.2.1 Interoception ... 3
1.2.2 Autonomous Functions ... 4
1.2.3 Food Consumption and Gustation ... 4
1.2.4 Nociception ... 5
1.2.5 Fear and Anxiety ... 6
1.2.6 Social Interaction ... 6
1.2.7 Addiction ... 7
1.2.8 Conclusion ... 7
1.3ANATOMICAL TRACING TECHNIQUES ... 8
1.3.1 Monosynaptic Retrograde Rabies Tracing... 8
1.3.2 Axonal Anterograde Viral Tracing ... 10
1.4OPTOGENETICS... 10
1.5FIBER PHOTOMETRY ... 11
1.6RODENT BEHAVIORAL TESTING ... 12
1.7AIMS OF THE STUDY ... 16
2 RESULTS ... 17
2.1WHOLE-BRAIN CONNECTIVITY MAP OF THE MOUSE INSULAR CORTEX ... 17
2.1.1 Starter Cell Characterization for RV and AAV Tracings ... 17
2.1.2 Whole-Brain Connectivity Map of IC ... 19
2.1.3 Pair-wise Correlations of Rabies and AAV tracings... 20
2.1.4 IC - Amygdala Connectivity ... 23
2.1.5 IC – Striatum Connectivity ... 25
2.2FUNCTIONAL CHARACTERIZATION OF POSTERIOR INSULAR CORTEX ... 27
2.2.1 Optogenetic Stimulation of IC Induces a Mixture of Aversive Behaviors ... 27
2.2.2 Bulk Calcium Imaging of IC Activity during Anxiety-Related Behavioral Tests ... 32
2.2.3 Optogenetic Manipulation of pIC during Anxiety-related Tests enables Bi-directional Modulation of Anxiety-like Behavior ... 35
2.2.4 pIC→CeA and pIC→NAcc projector neurons form non-overlapping populations .. 41
2.3FUNCTIONAL CHARACTERIZATION OF PIC→CEA AND PIC→NACC ... 44
2.3.1 Stimulation-induced Behaviors ... 46
2.3.3 Breathing Rate ... 48
2.3.4 Real-time Optogenetic Manipulation during EPM and EZM ... 49
2.3.5 Sucrose Preference Test ... 51
2.3.6 Feeding under Optogenetic Stimulation ... 54
2.3.7 Quinine Avoidance Test ... 55
2.3.8 Social Interaction ... 56
2.3.9 Lithium Chloride-induced Anorexia ... 57
2.3.10 Fear-induced Anorexia ... 58
2.3.11 Controlling for Back-propagating Action Potentials induced by Projection-specific Stimulations ... 60
3 DISCUSSION ... 62
3.1CONNECTIVITY OF THE MOUSE INSULAR CORTEX ... 62
3.2FUNCTIONAL CHARACTERIZATION OF PIC ... 66
3.2.1 Global pIC Manipulation... 66
3.2.2 pIC-to-Central Amygdala Pathway ... 67
3.2.3 pIC-to-Nucleus Accumbens Core Pathway ... 69
3.3LIMITATIONS OF THE STUDY ... 71
4 MATERIALS AND METHODS ... 73
4.1ANIMALS... 73
4.2VIRAL CONSTRUCTS ... 75
4.3STEREOTAXIC SURGERIES ... 75
4.4HISTOLOGY ... 76
4.5IMAGE PROCESSING AND DATA ANALYSIS OF TRACINGS... 77
4.6HEART- AND BREATHING RATE MEASUREMENTS IN ANAESTHETIZED MICE ... 80
4.7BEHAVIORAL TESTS ... 80
4.8FIBER PHOTOMETRY ... 85
4.9STATISTICAL ANALYSIS ... 88
5 REFERENCES ... 91
L
IST OF
A
BBREVIATIONS AND
A
CRONYMS
2P Two-photon
5-HT Serotonin
AAV Adeno-associated virus
aBLA Basolateral amygdala, anterior part ACC Anterior cingulate cortex
aIC Anterior insular cortex
AIP Agranular insular cortex, posterior
AP Anterior-posterior
APir Amygdalo-piriform transition area APs Action-potentials
Au2 Auditory cortex, secondary BDA Biotinylated dextran amine BDNF Brain-derived neurotrophic factor BLA Basolateral amygdala
BNST Bed nucleus of the stria terminalis
BW Body weight
CamKII Ca2+/calmodulin-dependent protein kinase CAV2 Canine adenovirus
CeA Central amygdala
CeC Central amygdala, capsular part CeL Central amygdala, lateral part CeM Central amygdala, medial part CGIC Caudal granular insular cortex ChR2 Channelrhodopsin-2
CM Centromedian nucleus of the thalamus
CNO Clozapine-n-oxide
CTA Conditioned taste aversion CTB Cholera toxin subunit B
D1R Dopamine 1 receptor
D2R Dopamine 2 receptor
DI Dysgranular insular cortex
DMT Dimethyltryptamine
DREADD Designer receptors exclusively activated by designer drugs
DV Dorso-ventral
ECT Ectorhinal cortex ENT Entorhinal cortex EnvA Envelope protein A
EPM Elevated-plus maze
eYFP Enhanced yellow fluorescent protein
EZM Elevated-zero maze
fMRI Functional magnetic resonance imaging
GABA Gamma-aminobutyrate
GFP Green fluorescent protein GI Granular insular cortex
GRIN Gradient index
HSV Herpes simplex viurs
IC Insular cortex
IPAC Interstitial nucleus of the posterior arm of the anterior commissure
LA Lateral amygdala
LFP Local field potential
LH Lateral hypothalamus
LiCl Lithium chlordie
LSD Lsyergic acid diethylamide LTD Long-term depression
LTP Long-term potentiation M1 Motor cortex, primary
MD Mediodorsal nucleus of the thalamus
MeA Medial amygdala
mIC Medial insular cortex
ML Medio-lateral
mPFC Medialprefrontal cortex
MSN Medium-spiny neuron
NAc Nucleus accumbens
NAcc Nucleus accumbens core
NBQX (2,3-Dihydroxy-6-nitro-7-sulfamoyl-benzo[f]chinoxalin-2,3-dion) NpHR3.0 Halorhodopsin 3.0
NTS Nucleus of the solitary tract
OF Open field test
OFC Orbitofrontal cortex
pBLA Basolateral amygdala, posterior part PBN Parabrachial nucleus
pIC Posterior insular cortex
Pir Piriform cortex
PRh Perirhinal cortex
RAIC Rostral agranular insular cortex ROI Region of interest
RTPA Real-time place aversion RTPP Real-time place preference
RV Rabies virus
S1 Somatosensory cortex,primary S2 Somatosensory cortex,secondary
SSRIs Selective serotonin reuptake inhibitors TMT Trimethylthiazoline
TVA Cellular receptor for subgroup A avian leukosis viruses
VP Ventral pallidum
VPL Ventral posteriolateral nucleus of the thalamus VPM Ventral posteriomedial nucleus of the thalamus
VPMpc Ventral posteriomedial nucleus of the thalamus, parvicellular part WGA Wheat germ agglutinine
1 I
NTRODUCTION
Affective as well as homeostatic states are strongly influencing ongoing behavior. These states arise from extended neural circuits through integration of internal and external stimuli1,2. In
particular, aversive states such as pain and negative moods tune behavior towards survival programs3. The insular cortex (IC) is a major cortical convergence site for internal and external
sensory information4–8. While rodent studies suggest an important role for the insula in processing
taste and bodily signals, human imaging studies have linked the IC with emotion processing and regulation5,8–10. With this evidence, our laboratory hypothesized that the insula serves as an
interface between bodily and emotional states. How this is mechanistically implemented is currently not known. In the following, I will briefly review the anatomy and function of the rodent insular cortex and introduce the main techniques that I used in this thesis.
1.1 Anatomy of the Rodent Insular Cortex
In animals that have a smooth-surfaced cortex, such as rats and mice, the insular cortex is located above the rhinal fissure on the lateral surface of the brain. The rodent insular cortex consists of several heterogeneous regions with distinct cytoarchitectural and connectional features11. In the
following, I will introduce the cyto- and chemoarchitecture as well as the connectivity of the rodent insular cortex
1.1.1 Cyto- and Chemoarchitecture of the Insular Cortex
The insula is comprised of (at least) three cytoarchitecturally distinct areas, according to the presence of granular layer 4 neurons. Based on this presence, the insula is separated into the granular (GI), dysgranular (DI) and agranular insula (AI). While the GI is organized in the classical six-layered prototypical cortical arrangement, the DI shows a progressive loss of the fourth layer.
Where this loss is complete, the AI begins and is composed of only three soma carrying layers (2/3, 5 and 6)12. AI, DI and GI form strong interconnections along the dorso-ventral and
rostro-caudal axes13,14. Along the rostro-caudal axis, the insula has been further divided into the anterior
insula (aIC), and the posterior insula (pIC). In rats, these are also sometimes referred to as rostral agranular insular cortex (RAIC) and caudal granular insular cortex (CGIC). The cytoarchitectural argument for a division into anterior and posterior part of the insular is less clear. The delineation of aIC and pIC are rather based on connectivity and functional responses.
Chemoarchitectural features of the insular cortex include the presence of acetylcholinesterase in the agranular insula, strongly reduced myelination in dysgranular and agranular insula, absence of SMI-32 immunoreactivity in the entire insula and reduced number of parvalbumin-positive interneurons in the agranular insula15. Further, several neurotransmitter receptors are expressed,
such as the D1- and D2 dopamine receptors16, ß-Adrenergic receptors17, cannabinoid
CB1-receptors18, serotonin receptors 5-HT
1A19, 5-HT2A19, 5-HT2C20,5-HT321 and possibly 5-HT4,5,6,722,
µ-, δ-, and κ-opioid receptors23 as well as nicotinic acetylcholine receptors24.
1.1.2 Connectivity of the Mouse Insular Cortex
The connectivity of the insular cortex has been characterized in rats and primates, but so far not as a comprehensive quantitative dataset. For the mouse, very few tracing studies exist25,26,
however, the Allen Brain Mouse Connectivity Atlas27 provides several anterograde axonal AAV
tracings from regions including IC. Yet, a brain-wide input map of the IC is completely lacking for the mouse. However, what studies have shown so far is that the IC receives information from external as well as internal senses via brainstem, thalamic and cortical inputs. These afferents are topographically organized leading to functional specializations, such as the ‘visceral insular cortex’, the ‘gustatory cortex’ (the primary taste cortex), or the insular auditory and somatosensory fields and potential many more28–31. Despite the specialization, it is important to
Further, the IC projects to BNST, the MD of the thalamus, the LH as well as to parahippocampal areas11.The IC is also heavily interconnected with other cortical regions11. Of interest in
emotion-regulation and decision-making is its reciprocal connectivity with prefrontal areas such as the OFC, mPFC and ACC. These regions are heavily implicated in executive-, cognitive- and emotional processing33. Further, the insular cortex, in particular the aIC has been shown to
strongly project to the ventral striatum in primates34,35 and rats36 and this connection has been
implicated in behavioral flexibility in reward contexts37,38.
Taken together, the IC and its subdivisions display a heterogeneous and extensive connectivity with limbic- sensory-, memory- and executive systems, which places the insula in a unique position for integration of multi-sensory, emotional and cognitive information. There is evidence that within the IC distinct connectivity profiles exists, suggesting different functional roles of the aIC and pIC.
1.2 Functions of the Insular Cortex
Human fMRI studies and electrophysiological recordings in primates and rats have revealed that the insula is involved in a plethora of situations but it is unclear what its exact role is39. In the
following, I will briefly introduce some key findings of the insular cortex stemming from both human and animal studies.
1.2.1 Interoception
Interoception – sensing the state of the body – has been correlated with activity in the insular cortex8,40. Anatomically, this is supported by inputs from nuclei of the thalamus which convey
information about the heartbeat, blood pressure, blood oxygenation levels, bowel movements, nociceptive stimuli, hunger, disgust, nausea, tickle, itch and many more5–7,10,40. The posterior
insula might be the first node in the brain, where all bodily information is integrated and monitored40. Subsequent processing towards more anterior parts of the insula is thought to further
integrate the emotional state, goals and memories40. This positions the insula to detect salient
stimuli regarding bodily integrity and homeostasis. Evidence supporting this, comes from experienced mindfulness meditators, which display an increased activation as well as an increased grey matter volume of the right anterior insula41,42. Another very interesting finding that
corroborates the role of IC in interoception and self-awareness came from the group of Robin Carhart-Harris in 2016. They wanted to image human brains under the influence of the
hallucinogenic compound LSD (lysergic acid diethylamide). After administration of LSD (i.v., 75 µg) the healthy volunteers were subjected to fMRI scans. Strikingly, the functional connectivity density of the insular cortex was highly correlated with self-reports of ego dissolution. This indicates a higher functional connectivity of the insular cortex with other cortical areas under the influence of LSD, possibly explaining out-of-body experiences or delusions to transform into an animal (lycanthropy)43. Another potent psychedelic substance
Dimethyltryptamine (DMT), which is the main ingredient in the ancient shamanic Amazonian brew “Ayahuasca”, has strong anti-depressant and ego-dissolving effects. Interestingly, healthy and depressed subjects receiving Ayahausca and undergoing single photon emission tomography (SPECT) displayed a strong modulation of insular cortex activity44,45. Almost all of the other
functions of IC that I will discuss below might eventually be related back, at least in part, to the interoceptive role of the insular cortex.
1.2.2 Autonomous Functions
In addition to sensing bodily aspects – i.e. interoception – the insular cortex can also affect autonomic functions directly, such as regulating the heartbeat, blood pressure or bowel movements46–50. It is unclear, via which pathways these effects could be mediated, but likely
candidate projections include those to the BNST, CeA, LH or PBN. Studies in rats uncovered adjacent ‘pressor’ and ‘depressor’ sites within the pIC by electrical microstimulation in anesthetized animals51. Stimulation of these sites displayed opposing effects on the heart rate and
blood pressure. This implies the existence of separated and antagonistic circuits very close to each other within the insula. This makes the insula an interesting structure in the etiology of psychosomatic disorders.
1.2.3 Food Consumption and Gustation
Multiple neuroimaging studies in humans and experiments in rodents revealed the contribution of the insula in gustation52,53, for visual food cues54,55, or olfactory food cues56, but also during
craving for food57. Ultimately, eating is a multimodal experience composed of taste, smell and
texture58. These sensory cues are thought to be combined within the anterior insula, as suggested
by overlapping activation of parts of the anterior insula after independent stimulation with odor and taste cues59,60. In addition, the viscosity and texture of food is correlated with activity in the
aIC and mIC61,62. An important role of the assessment of the consumed food is how one’s body
feels after having eaten. This relates to the interoceptive role of the insula, as described above40.
Another supporting evidence that the IC is implicated in gustation and consummation of food is insular pathophysiology in eating disorders. A meta-analysis found a decreased mIC activity in overweight compared to lean subjects63. The authors hypothesize that a maladaptive
interoception, causing a misinterpretation of appetite signals of the gut, could increase food intake in obese subjects.
Conditioned Taste Aversion
Conditioned taste aversion (CTA) is an acquired avoidance of a novel taste (= conditioned stimulus) if it has been previously paired with the sensation of visceral malaise or sickness (= unconditioned stimulus)64–66. In animal models, this can be reliably induced by pairing a novel
taste with the administration of lithium chloride (LiCl)64. The mechanisms mediating conditioned
taste aversion have been extensively studied in rats and mice and revealed, that lesions or pharmacological silencing encompassing the insular cortex prevent the acquisition, storage and expression of conditioned taste aversion to novel but also to familiar tastes67–70. Further studies
showed the involvement of cholinergic and dopaminergic signaling in regulating insular plasticity during CTA71,72. Recently, Lavi et al. found that conditioned taste representations within the IC
selectively shift to BLA-projecting IC neurons73. These results suggest the presence of
valence-specific neurons with a direct influence on the expression of taste aversion. Through learning, sensory stimuli are assigned to such valence-specific neurons, or in the case of innate preference for sucrose or avoidance of bitter taste, may be hard-wired.
To summarize, the aIC and mIC seem to be multimodal integration hubs for the processing of food-related stimuli and maladaptive responses in these areas could underlie or promote obesity. Further, the insular cortex plays a pivotal role in the acquisition, storage and recall of conditioned taste aversion.
1.2.4 Nociception
Noxious somatosensory stimuli often activate the insular cortex, which has been suggested to play an important role in pain processing74–76. The pIC responds to the somatosensory aspects of pain,
while the aIC has been shown to mediate its affective features77. Accumulating evidence
implicates anatomical and functional alterations in the IC in the expression of chronic pain and with cognitive and affective disorders78. Interestingly, alterations in insular grey matter volume
due to chronic pain could be partially reversed after successful therapy of trigeminal neuralgia79.
The role of the insula in pain is again closely intertwined with interoception, particularly, in the shaping of pain perception. Here, the aIC integrates multimodal information including nociception to give rise to the affective dimensions of pain.
1.2.5 Fear and Anxiety
Human imaging studies as well as animal studies have revealed a role for the insular cortex in representing positive and negative emotions, such as anger, sadness, fear and anxiety, disgust, happiness or joy, trust, surprise, as well as social emotions39,40,80. Emotional stimuli of various
modalities, such as touch, vision or olfaction, elicit IC responses. Thus, a growing number of studies suggest an important role for the IC in fear and anxiety10,39,40,81. Human fMRI experiments
detected a functional and structural connectivity of the IC with other limbic brain areas, in particular with the amygdala and the strength of this connectivity has been shown to correlate with the anxiety levels of healthy subjects81,82. In animal studies, electrophysiological recordings
during fear-conditioning experiments revealed a population that responds to freezing and extinction within pIC83. Furthermore, irreversible lesion experiments as well as temporary
pharmacological inhibition of the rat IC revealed a role of the IC in consolidation of learned fear84,85. Interestingly, the IC seems to be also involved in the learning of safety cues, which signal
the absence of the unconditioned stimulus86,87. Therefore, both fear increasing as well as fear
reducing circuits were found within the IC.
1.2.6 Social Interaction
Social animals need to be able to recognize and transmit social cues that facilitate and enable social interaction88. This is achieved by the social-decision-making network89, which is thought
to include the insular cortex, as lesions of IC in humans and rodents results in deficits in emotion recognition and empathy90–93. This is further supported by changes in insular cortex activity and
connectivity in autistic subjects94–96. Of particular interest in the context of social interaction is
the oxytocin-releasing projection from the paraventricular nucleus of the hypothalamus to the agranular insula97. A recent study in rats has shown that oxytocin-release in the IC orchestrated
approach or avoidance to stressed or non-stressed conspecifics98. A follow up study by the same
lab identified the projection of the IC to the NAc to modulate social approach99. As social
interaction seems to be based on recognizing one’s own bodily and emotional state and integrate it with those of a conspecific, the interoceptive function of insular cortex make it a prime candidate for modulating social interactions.
1.2.7 Addiction
Addiction, also defined as “substance use disorder” in the DSM-5 (Diagnostic and Statistical Manual of Mental Disorders, 5th edition), is a mental disease that is defined by compulsive drug
use and seeking despite negative consequences100. Furthermore, frequent relapse behavior is
characteristic of addictive behavior100. Functional imaging studies in addicted human subjects
have repeatedly and across several drugs of abuse detected correlations of insula activity upon drug consumption and drug seeking101. In addition, grey matter volume of the insular cortex is
reduced in cocaine addicts102. In a very interesting analysis, Naqvi et al. report that stroke patients
with lesions including the insular cortex, display an immediate and long lasting cessation of smoking in heavy smokers. These patients reported a lack of conscious bodily urge to smoke after the stroke. This was not the case in stroke patients where the insular cortex was not affected by the stroke103. Rodent studies have further supported the role of the insula in the formation of
amphetamine-induced conditioned place preference, a model for drug seeking behavior4. Another
study in rats has found that aversion-resistant intake of alcohol is sustained by glutamatergic projections from the insula and mPFC to the nucleus accumbens core104. Further, intra-venous
self-administration of nicotine in rats was reduced upon infusion of a dopamine D1-receptor antagonist into the insular cortex, showing the importance of the dopamine system in the insular cortex for drug rewards105.
These studies in rodents and humans suggest two modes of action by which the IC could influence addictive behavior. First, the acute urge to seek and consume a drug – i.e. craving, seems to be manifested via the IC. Second, aversion-resistant drug consumption might be caused through maladaptation of insular circuits that would otherwise terminate ongoing drug use.
1.2.8 Conclusion
As introduced above, the insular cortex has been implicated in a broad variety of functions, from taste over pain to addiction. Although at first many functions seem to be unrelated to each other or incoherent, a general finding shared across different species does emerge. The insula serves as a hub where various sensory, emotional and memory systems are integrated. Through its afferents, the insular cortex is in a perfect position to monitor the internal as well as external environment of an organism. Further, the insula adapts its predictions of future internal states through learning, as is evident by conditioned taste aversion. It therefore plays an important role in capturing the valence of internal or external stimuli, which might be maladapted, e.g. in drug or food addiction. These findings could explain why insula integrity is necessary for learning, emotion regulation,
and decision-making. In addition, evidence points to the fact that the insula is part of a salience detection network, which prioritizes the processing of acutely relevant stimuli. How exactly this is implemented on a mechanistic or cellular level is currently unknown. We also have little information about the function of specific output pathways or how such circuits might differ in their connectivity.
1.3 Anatomical Tracing Techniques
1.3.1 Monosynaptic Retrograde Rabies Tracing
Non-viral tracing techniques, such as Cholera toxin subunit B (CTB), biocytin, biotinylated dextran amine (BDA), wheat germ agglutinin (WGA), fluorogold, retrobeads, and others, were invaluable in assessing the connectivity of the brain. However, these methods cannot account for cell-type specific differences of connectivity and they do not provide information about monosynaptic inputs.
Retrograde viral tracing techniques based on neurotropic viruses, such as herpes-simplex virus (HSV), canine adenovirus type 2 (CAV2) and rabies and pseudo-rabies virus (RV, PRV), exploited the ability to traverse (multi-)synaptic pathways while being amplified at every step. These methods required at least bio-safety level 2 laboratories and for multi-synaptic tracings, the timing of termination of the tracings determined how many synapses were traversed. As these viruses hijack the endogenous transport machinery, there can be a marked difference of how many synapses were traversed, depending on the physical length of the axons.
A major advance in rabies-based circuit tracing was the deletion of the glycoprotein “G” from the rabies genome, yielding RV∆G, which is restricted to a single trans-synaptic retrograde traverse (i.e. monosynaptic)106. In 2007, Wickersham et al. developed a cell-type specific monosynaptic
retrograde rabies tracing technique by pseudotyping RV∆G with the envelope protein A (EnvA) of the avian leukosis virus. This restricted rabies infection to cells that expressed the avian receptor TVA107. By expressing the TVA receptor in a Cre-dependent manner, it became possible,
After retrogradely traversing from the starter neuron to the presynaptic partner, which must not express the G protein, a further trans-synaptic jump is not possible, thus only labeling direct monosynaptic input neurons.
This novel and powerful strategy was quickly adapted in the systems neuroscience field and has been used to trace from single neurons108, from projection-defined neurons109, from newborn
cells110, and most frequently from genetically defined cell types using Cre driver lines111–116.
Although various approaches are conceivable, most frequently, a Cre-dependent AAV construct coding for the G protein and the TVA receptor is injected into a brain region of interest in a Cre-driver mouse. After a period of 2-3 weeks allowing for sufficient expression of the gene products, a GFP-expressing EnvA-pseudotyped RABVΔG is infused in a second stereotaxic surgery into the same region in order to infect the TVA- and G- expressing neurons (see Figure 1).
Figure 1. Schematic of cell-type specific monosynaptic rabies tracing. 1) A mixture of two
Cre-dependent AAV helper constructs is injected into the target region. After 2-3 weeks, a sufficient expression of TVA and G should be achieved. 2) Then, the genetically modified rabies virus, lacking the gene for the G protein and pseudotyped with the EnvA capsid is injected into the same area. Neurons that were infected with both, the two helper constructs and the rabies virus represent the “starter cells” (yellow). The brain-wide presynaptic partners (input neurons) to these starter neurons will be retrogradely infected by the rabies virus and express GFP.
1.3.2 Axonal Anterograde Viral Tracing
In order to visualize brain-wide projections of neurons, neuroanatomists started using recombinant adeno-associated virus (AAV) expressing a fluorophore117. AAV vectors are based
on non-pathogenic and defective human parvovirus and they cannot replicate without a helper virus. The major advantage of recombinant AAV (rAAV) is its very low toxicity and stable long-term expression of gene products118. Aside from anatomical tracing studies, this makes the AAV
system an attractive technique to deliver gene therapy118.
The combination of the Cre-LoxP technology with rAAVs enables cell-type specific anterograde axonal tracings. To achieve this, the recombinant gene of interest, which is cloned into the AAV sequence, is flanked by two LoxP sites, rendering it dependent on the Cre-recombinase for efficient gene expression. In neurons that do not express Cre, very low levels of the gene product are observed. For anatomical studies, a fluorescent reporter protein such as the green fluorescent protein (GFP) are usually employed. GFP expression fills the entire neuron after a few weeks enabling the brain-wide visualization of its axons. This allows for a qualitative and quantitative measurement of innervation patterns of given cell-types.
1.4 Optogenetics
The development of optogenetics allowed neuronal manipulation in-vitro and in-vivo with a high-temporal and cellular precision119. The principal idea of optogenetics is to exploit the
light-sensitivity of microbial opsins, to control ion-flow in neurons. Opsins are seven-transmembrane proteins, which exist in many variants, differing in their photocurrents, ion-specificity, kinetics and expression profiles. A major breakthrough was achieved in 2005 in Karl Deisseroth’s lab at Stanford, which performed experiments expressing the algal protein Channelrhodospin-2 (ChR2) in cultured neurons120. They could reliably control neuronal firing in a millisecond-timescale by
applying blue light (473 nm). In the last 14 years many new genetically engineered and optimized opsins became available, leading to a large repertoire of specialized tools to either activate or inhibit neuronal spiking or gene expression, on varying time-scales.
The in-vivo use of optogenetics led to an unprecedented advance in understanding the function of neural circuits. By expressing activating or inhibiting opsins in genetically defined neurons and implanting optic fibers over these neurons to deliver light, we gained mechanistic understandings of circuits and their implications in behavior. An extensive introduction of current advances in optogenetics is out of scope of this thesis and I refer to the many excellent reviews and protocols
However, very relevant to this thesis is the projection-specific manipulation of axon terminals
in-vivo. This strategy specifically manipulates long-range projections of opsin-expressing neurons
in a target region, by placing the optic fiber that delivers the light over that target region121. This
enables the functional characterization of a specific pathway, without affecting the cell bodies of the opsin expressing neurons.
Nevertheless, there are caveats when applying optogenetics to study neural circuits in-vivo. Commonly, fixed stimulation frequencies, ranging from 1-60 Hz, have been used to stimulate large populations of neurons for one to many hundreds of seconds. Obviously, this pattern of activation does not resemble physiological firing of neurons and is highly artificial. Especially when assessing the behavioural outcome of these stimulations, one needs to consider that the observed behaviors might not occur in such a fashion in nature123. However, compared to
chemogenetic manipulation124 or electrical microstimulation, optogenetics is still the most
temporally precise tool to study genetically-defined neural circuits.
1.5 Fiber Photometry
Measuring neural responses via calcium imaging using genetically encoded calcium indicators, like the GCamP125 family, played an important role in the functional characterization of neural
circuits in neuroscience. Two-photon imaging126,127 enabled the recording of many hundreds of
single neurons in the layers 2/3 of cortex, and with recent advances in multi-photon imaging, neurons can be recorded up to >1 mm deep into cortex128. However, these costly and delicate
microscopes usually require the mice to be head-fixed or anesthetized. Single-photon calcium imaging of deep brain structures during freely moving behavior became possible with head-mounted miniature endomicroscopes129, however, the weight (>2g) of such a microscope can be
as much as 10-20 % of the bodyweight of a mouse. Further, to image deeper areas, the use of graded-index (GRIN) lenses is necessary, but the large diameter of these lenses (>500 µm) causes substantial damage to the brain. New two-photon miniature microscopes are being developed and show promising results130, yet, the issue with the GRIN lenses and weight is not solved.
Alternatively, fiber photometry provides a comparably easy solution to record the bulk neuronal activity of genetically-defined populations131,132. It makes use of the same multimode optical
fibers that are also used for optogenetics and are implanted above the GCamP-expressing neurons. Through the same fiber, excitation light is delivered to the neurons and the resulting fluorescent signal of the calcium indicator collected. Thus, the mobility of the mouse is only mildly affected and more complex behaviors, such as social interactions, can be studied. New multi-channel fiber photometry systems have emerged that allow simultaneous recordings of two133 or more sites134.
A disadvantage of fiber photometry is the loss of single neuron resolution, compared to microendoscopy. If the signal is recorded from a heterogeneous population, like in cortex, responses from sparse or rarely active populations might be underrepresented and those that fire in synchronicity or very frequently might be overrepresented. This issue can be tackled by further refining the cell-type specificity via projection targets or intersectional genetics. Nevertheless, the bulk activity can be a first step in characterizing the dynamics of a brain region.
1.6 Rodent Behavioral Testing
Behavior has been defined by Tinbergen as “the total movements made by the intact animal”135.
In order to understand how the brain generates a particular behavior, we need to identify the putative neural circuits that ultimately trigger the movement123. In the case of emotion-related
behaviors, exposing rodents to specific environments can trigger behaviors that resembles anxiety or fear in humans. Manipulating the activity or integrity of brain areas during such behavior has helped us identify and partially understand the neural circuits involved in motivated behavior. It is important to keep in mind that almost all commonly used behavioral tests use a reductionist approach to make a very specific set of behaviors amenable to neuroscientific analysis. A single behavioral test cannot and does not try to address all multifaceted and highly complex naturalistic behaviors at once. Instead, behavioral tests can be designed in such a way that they model very specific domains of behavior that can help us to eventually tackle the underlying causes for maladaptive disorders, such as anxiety, depression, or addiction. In the following, I will briefly introduce the behavioral tasks that will be used in this thesis.
Anxiety-related Behavioral Tests
Open Field Test
The open field test (OF) has been used in rodent behavioral research since 1934 an is one of the most frequently used instruments in animal psychology136. It is so popular, because it is extremely
simple to perform, does not require any training of the animal and there is a consensus on the interpretation of the results. Rodents are placed into a round or rectangular apparatus that is enclosed with walls and allowed to explore the novel environment. Originally, Hall developed the OF in 1934 to count the number of fecal boli of rats as an index of timidity137. Over time,
many more parameters have been added to the analysis of the OF. Today, most frequently assessed parameters are locomotor activity and the ratio between “time spent in the center” vs. “time spent in the periphery”. While stimulants such as amphetamine will strongly increase the
distance travelled in a OF, sedatives reduce locomotor activity136,138. As rodents show an innate
avoidance of brightly lit and open spaces, they will prefer to move along walls (thigmotaxis) and spend less time in the exposed center zone. Interestingly, the anxiolytic benzodiazepine diazepam increases the time spent in the center zone without affecting the total distance travelled139. Since
then, this constellation of results is considered to be indicative of an anxiolytic effect. Other measures related to the anxiety state of an animal include the stretched-attend position and rearings136. Recent advances in quantification and characterization of behaviors by machine
learning techniques enable sub-second classification of behaviors and provide novel ways of analyzing OF data140.
Elevated-Plus Maze
The elevated plus maze (EPM) was first introduced in 1984 and relies on rodents’ preference for darker and enclosed spaces and innate avoidance of heights and open spaces141. The EPM assesses
the internal conflict of a rodent to avoid predation on one hand and to explore its surroundings for food and mating partners on the other hand. While the closed arms provide protection, the open arms expose the rodent to height and an open space, but also a potential – but ultimately not presented – reward. To perform the test, the animals are placed in the center of the maze facing a closed arm and their position in the maze is tracked for 5-10 min. The main readout for this behavioral task is the ratio of the time spent on the open arms vs. the closed arms, but parameters such as distance travelled, number of entries, number of stretched-attend postures, freezing or fecal boli might be of value142.
In general, mice will spend much more time in the closed arms and depending on the conditions (mouse strain, age, sex, handling, lighting, satiety, pre-exposure to novel environment, circadian rhythm, housing conditions, to name a few) between 5 and 30 % on the open arms.
What helped the EPM to gain traction in the preclinical research community was that compounds that display anxiolytic effects in humans, such as benzodiazepines, lead to an increase of time spent on the open arms, while leaving total distance travelled unaffected. In contrast, compounds that increase anxiety in humans also decreased the time spent on the open arms in rodents143,144.
This grants the EPM a high construct validity.
Nevertheless, criticism as a translation model of anxiety arose, as the EPM is only predictive for anxiolytic drugs acting on the GABAergic system, while newer compounds, e.g. 5-HT1A
antagonists or selective serotonin reuptake inhibitors (SSRIs), which have clinical evidence of anxiolysis in humans, do not influence the time spent on the open arms145. Yet, the EPM and its
variants, such as the elevated zero maze (EZM) remain the most popular anxiety-related tests to date.
Reward-related Behavioral Tests
Real-Time Place Preference or Avoidance
The Real-Time Place Preference (RTPP) or Real-Time Place Avoidance (RTPA) test is a modification of the classical conditioned place preference test (CPP)146, but adapted to
optogenetic manipulation. First, I will briefly introduce the CPP, which is an established test in neuropharmacological research to assess the motivational effects of psychoactive compounds. In essence, one compartment of a two-chambered behavioral apparatus is (repeatedly) paired with the administration of a compound of interest. Should the substance possess rewarding or habit-forming properties, animals will spend more time in the chamber that has been paired with the administration of the substance, as compared to the chamber where this pairing never occurred. In contrast, should the substance elicit aversive properties, the animals will avoid the paired chamber.
While the CPP assesses the memory of an outcome, the RTPP/RTPA enables the direct interrogation of the acute valence of an optogenetic stimulation. Analogous to the CPP, one of the two chambers is designated as the paired chamber, where optogenetic manipulation will occur. This is achieved by tracking the position of the animal in real-time and thereby constructing a closed-loop circuit which controls the laser emission. In case of a rewarding or pleasurable effect of optogenetic manipulation, the animals exhibit a real-time place preference. In contrast, aversive effects mediated by optogenetic manipulation should lead to reduced time spent in the stimulated chamber – a real-time place aversion. The RTPP has been a valuable tool in assessing positive or negative behavioral effects of optogenetic manipulations9,147–149.
Sucrose Preference Test
The sensitivity to a hedonic stimulus, for example to a sucrose solution, can be tested with the sucrose preference test. The test measures, how much rodents prefer a sucrose solution over water. Under normal circumstances, most strains of rats and mice strongly prefer the sweet sucrose solution; however, when the animals become anhedonic, for example by exposing them to chronic stress or other models of depression, they reduce that preference. Therefore, the sucrose preference test is a very simple but yet powerful test to approximate the sensitivity to rewards. A major advantage of this assay is its technical simplicity. On the other hand, low reproducibility and high variability can be disadvantages. Interestingly, a study showed that the common practice of handling mice by their tails can decrease sucrose consumption and licking bouts150. Therefore
careful attention should be paid to adhering to a strict protocol when performing the sucrose test151. Important confounders are the duration of the test, the protocol for food and water
As anhedonia presents a major symptom in depressed human patients but also in patients with schizophrenia, the sucrose preference test has been very frequently used in preclinical translational research151. Interestingly, administration of antidepressants, such as the SSRIs or
tricyclic antidepressants, can restore the sucrose preference in rodent models of depression.
Social Interaction Test
Social motivation to interact with conspecifics is a powerful drive of humans and deficits in social interactions, as seen in autism spectrum disorder, can be very detrimental152. Mice are also a social
species and exhibit many diverse social behaviors, such as reciprocal social interactions, social play, parenting, mating or aggressive behaviors153. Researchers developed several behavioral
assays to assess certain aspects of social behavior in mice153. One of the most commonly used
tests to measure social reward processing is the three-chamber social interaction test154. Here, the
animal can freely explore a maze in which one compartment of the chamber harbors another conspecific and the other compartment a non-social object. Under normal conditions, mice spend more time with the social stimulus than with a novel object. However, chronically stressed – i.e “depressed” – animals reduce their social interaction time153. Therefore, the social interaction test
is another measure of the hedonic state of an animal, but not reliant on taste and thus an interesting complement to the sucrose preference test.
1.7 Aims of the Study
While anatomical investigations in diverse species highlight that the insular cortex is one of the most complex anatomical hubs in the mammalian brain50,155–158, to date there is no comprehensive
connectivity map of the mouse insular cortex.
Therefore, the first goal of this thesis was to map the input and output connectivity of IC
excitatory as well as IC inhibitory neurons, by performing monosynaptic retrograde rabies tracings and anterograde axonal tracings of the mouse IC. I divided the IC into three equally large subdivisions, which I termed anterior, medial, and posterior insular cortex (aIC, mIC, and pIC, respectively). The resulting data set would serve as an anatomical framework to guide functional studies.
As the pIC has been hypothesized to be the entry point into the IC, the second goal was to
functionally analyze the role of the pIC in emotion-related behavioral tasks. Previous work from the Zuker lab9,30,148 characterized the pIC in the processing of bitter taste, but evidence suggests
that the pIC plays an important role beyond taste recognition4,86,159,160. To clarify and further
extend the idea that the pIC is involved in emotion-related processes, I employed optogenetic manipulations of the pIC during anxiety-related and consumption-related behavioral tasks. Based on the anatomical findings I would obtain from the first goal, as a third goal, I wanted to
then functionally characterize segregated IC populations that would project to interesting targets of the limbic system, such as the amygdala, mPFC, ventral striatum or ventral hippocampus and try to disentangle their specific roles.
2 R
ESULTS
2.1 Whole-Brain Connectivity Map of the Mouse Insular
Cortex
2.1.1 Starter Cell Characterization for RV and AAV Tracings
I first checked the quality and distribution of the starter cell populations for both rabies virus (RV) tracings and adeno-associated viral (AAV) tracings with a CellProfiler pipeline that was created by myself with the help of T. Gaitanos, a further lab member (Figure 2a, b and Methods). For
the RV tracings, I scored neurons as starter cells when they were double positive for eGFP (i.e. RV+) and mCherry (i.e TVA+). For AAV tracings, starter neurons were detected as eYFP-positive cell bodies. This provided me with the total number and location of all starter cells, which are depicted in Figure 2c. The starter cell populations for the three IC subdivisions were separated
and mostly non-overlapping for both AAV and RV tracings. In a few cases, a small amount of starter cells were found outside the IC (e.g. in Pir, S1 and S2, or M1). A comprehensive table containing information for all animals used for the anatomy study can be found in Table 1
Figure 2. Characterization of starter neurons for RV or AAV tracings. a) Starter cell identification pipeline for
Rabies tracings. 1. High resolution image of a representative section at the injection site in the IC. Starter cells are double-labelled with TVA-mCherry and RV-GFP, and appear yellow. Scale bars 200 µm (main image), 50 µm (inset). Number of starter cells was identified in an automated fashion using Cell Profiler. First RV+ cells were identified (image 2, yellow cell outlines) from the GFP image, then RV+ cells that also contain mCherry-TVA were identified from the mCherry signal (image 3, red rings within yellow RV+ cell outlines). Double labelled cells are counted as starter cells. b) Starter cell identification pipeline for AAV tracings. 1. Representative epifluorescent image of YFP-labelled AAV starter cells. Scale bars 200 µm (main image), 50 µm (inset). 2. YFP-positive cells were identified in an automated manner using Cell Profiler. c) Schematic illustration of the lateral view of the IC including distances from Bregma (top panel) and heatmap showing average starter cell distribution for each tracing strategy at each specific IC target (bottom panels). The three IC target subdivisions were mostly non-overlapping, and only a minimal percentage of cells were detected in the Motor and Sensory Cortex (M/S), or Piriform Cortex (Pir) neighboring the IC. n = 3 mice per injection site/tracing strategy. Heatmap intensity scale is the same for all three IC target subdivisions. Regions absent at
2.1.2 Whole-Brain Connectivity Map of IC
To obtain an overview of the macro-scale IC connectivity, I created a brain-wide input and output connectivity map of excitatory and inhibitory neurons of the aIC, mIC and pIC. I segmented the data into 17 major brain regions that connect to the IC (Figure 3). As there was an order of
magnitude difference between cortical and subcortical regions, I separated the data into these two categories for easier side-by-side comparison (Figure 3).
Figure 3. Whole-brain input and output map of the mouse insular cortex. Comparison of inputs
to excitatory and inhibitory IC neurons (left) and outputs of excitatory neurons of the IC (right) of all three IC subdivisions (aIC, red; mIC, green; pIC, blue) across the 17 major brain regions that displayed connectivity. Region values are given as percentage of total cells (RV) or of total pixels (AAV). Data is shown as average ± SEM. n = 3 mice per condition. Top panel shows cortical connectivity, bottom panel shows subcortical connectivity.
There were no substantial differences between the inputs to excitatory and inhibitory neurons, irrespective of the IC subdivision from which I traced. There was also no region, which solely innervated one of the three IC subdivisions. Yet, there were quantitative differences depending on region and IC subdivision. For example, there were around two times more inputs from the sensory cortex to the pIC (41 ± 11% of total excitatory input connectivity) in comparison to the other IC subdivisions (20 ± 5% and 23 ± 7% for mIC and aIC, respectively). In contrast, the motor cortex provided the majority of inputs to the aIC. In subcortical regions, the amygdala and thalamus provided most inputs to the IC.
While there was a strong bi-directional connectivity with other cortical regions, some subcortical regions showed a strong directional bias, favoring either inputs or outputs. For example the striatum, which did not project to the IC, but was heavily innervated by it. In particular, the aIC provided the strongest projection to the striatum (32 ± 6%, as compared to 9 ± 1% and 11 ± 2% for the mIC and pIC, respectively).
I also analyzed the data at a more detailed level, analyzing 75 brain areas separately. The resulting comprehensive data describe percent of total input to excitatory and inhibitory neurons, as well as outputs from excitatory neurons of the aIC, mIC and pIC. The tables containing these data can be found in the Appendix (see Appendix 1, 2, 3).
2.1.3 Pair-wise Correlations of Rabies and AAV tracings
Throughout my analysis, for both inputs and outputs, I noticed differences between the aIC, mIC and pIC. To test whether my observations could be confirmed by an unbiased statistical analysis, I correlated all input tracings (including inhibitory and excitatory mouse lines) to each other, as well as correlating all output tracings to each other. To do this, I calculated the pairwise correlation coefficients from the data of the 17 major brain regions (Figure 4a,b), but also at the highest
resolution with all 75 regions of interest that have been used and applied a hierarchical clustering method (Figure 4c,d).
Comparing the data at the level of 17 major brain regions was chosen to detect a global and striking difference between major brain systems. This analysis showed that overall, the input-input comparisons displayed rather high correlation coefficients (Figure 4a, average correlation
coefficients of 0.7 ± 0.16), indicating that across aIC, mIC and pIC for both excitatory and inhibitory neurons, inputs were scaling very similarly. The hierarchical clustering almost perfectly separated tracings from aIC (red labels and red dendrogram) of those from mIC or pIC, which were intermingled. The mIC and pIC cluster was further divided into two sub-clusters, however, this was neither due to injection site nor due to the genotype.
Surprisingly, when taking every ROI into account (Figure 4c, 75 brain regions) and performing
the same analysis, the average of all correlations of the rabies tracings did not change (0.7 ± 0.17). This suggests that the main differences between the tracings are already evident at a higher level of the 17 major brain systems. Again, the hierarchical clustering grouped all except one tracing into the aIC cluster, while intermingling mIC and pIC tracings in the other cluster.
Figure 4. Hierarchical clustering separates aIC from mIC and pIC for inputs as well as outputs.
This is the case when correlating averaged data of 17 major brain regions (a, b, higher hierarchy) and when correlating the lowest level of all 75 brain regions (c, d). Matrices of hierarchically clustered pair-wise correlation coefficients (Pearson’s) of animals inputs vs. inputs a), c) or outputs vs outputs b), d). The pair-wise correlations were performed on the data organized into 17 major brain regions (see Figure 3), or on all 75 regions (see Appendices 1-3). Far left gradient
bar (green hues) indicates the center of the starter cells, ranked relative to every mouse in the dataset. Note for both input and output correlations, a clear cluster forms from the aIC-targeted
animals (top left boxed sections and red-colored dendrograms), whereas the mIC- and pIC-targeted animals intermingle in a second cluster (larger boxed areas). Interestingly, the clustering algorithm did not separate excitatory (CamKIIα) from inhibitory (Gad2) rabies tracings.
Analyzing the pairwise correlations for the AAV tracings at the 17 major brain region level, yielded an average correlation coefficient of 0.45 ± 0.28 (Figure 4b). This indicated that the
outputs were still positively correlated with each other, but less so compared to the inputs vs. input correlations. Again, all three aIC AAV tracings were grouped into the same cluster, while AAV tracings from mIC and pIC were grouped together, except one outlier (mIC).
Repeating the analysis including all 75 regions yielded a very similar average correlation coefficient of 0.46 ± 0.27. Again, this implies that most of variance is already captured with the 17 major brain regions.
To summarize, for both inputs and outputs, after applying hierarchical clustering, two distinct clusters formed, dividing the aIC animals from the intermixed mIC- and pIC mice. Indeed, for both input vs. input correlations as well as output vs. output correlations, the mIC and pIC tracings were interleaved, indicating a high degree of similarity between the medial and posterior regions of the IC. Neither the anterior-posterior coordinate of the starter cell population (left columns, green gradient), nor the cell-type from which I traced, could explain the sub-cluster within mIC and pIC targeted animals.
The striking difference of input- and output-patterns of aIC connectivity imply a functional difference as compared to mIC and pIC. In particular, for the efferents, a major difference arises from the strong projection to the striatum, and to a lesser extent, to the motor cortex (Figure 3).
2.1.4 IC – Amygdala Connectivity
The IC and amygdala are heavily interconnected through direct but also indirect pathways, some of which are reciprocal13,161–163. Both regions are heavily implicated in emotional processing and
I therefore focus on the IC-Amygdala connectivity in this section.
As expected, only cortex-like sub regions of the amygdala, such as the BLA or APir provided inputs to the IC subdivisions, whereas they were completely absent from striatum-like nuclei (CeA and MeA) (Figure 5a, Figure 6a). Interestingly, afferents from APir represented the only
case where meaningful differences between tracings from excitatory and inhibitory neurons were detected.
While inhibitory neurons of the aIC received very few inputs from the APir, excitatory neurons received considerably more (0.07% vs 1.50 %). In addition, I observed a pIC-to-aIC gradient regarding the APir inputs to the inhibitory neurons, with high input to pIC (>3% connectivity in the pIC) and very low to aIC.
Interestingly, only small variations were detected when comparing inputs from the amygdala to excitatory neurons between aIC, mIC and pIC, with the mIC receiving slightly more innervation. This finding was further supported by the analysis of input density, which revealed a broader rostro-caudal spread and higher density of mIC-projecting aBLA, pBLA and APir neurons (Figure 6b).
Regarding the outputs from the IC, all amygdalar regions were innervated except the MeA (Figure 6a). I found two opposing amygdala innervation gradients along the anterior-posterior
axis of the IC. For instance, pIC strongly projected to the CeA, APir and pBLA, which confirms
Figure 5. IC-amygdala connectivity. a) Coronal sections depicting the amygdala with its
subregions. Distances are provided as anterior-posterior positions relative to Bregma. b) Representative images from excitatory inputs (top row, eGFP-expressing nerons) and outputs (bottom row, eYFP-positive neurons). Different Bregma levels are shown for each IC target site, as indicated on the images (-0.9 mm, -1.2 mm, -2.0mm). Scale bar = 200 µm.
retrograde CTB studies performed in rats and mice, respectively164,165. In the second part of this
thesis, I functionally analyzed the pIC→CeA pathway using optogenetics and could reveal an important role in anxiety-related behavior (see Chapter 2.3). On the other hand, the aIC mainly
projected to the aBLA. Recently, this aIC-to-aBLA projection has been shown to elicit appetitive responses9. Further, the aIC did not innervate more posterior basolateral regions (pBLA), which
can be seen in the innervation density plot in (Figure 6b). In addition, the aIC did not project to
any other amygdaloid nucleus except from a sparse projection to the EA.
To summarize, there is a clear difference between the IC subdivisions when comparing outputs to the amygdala. Interestingly, the inputs to all subdivisions were quantitatively very similar.
Figure 6. Quantification of Insula-Amygdala connectivity. a) Comparison of excitatory and
inhibitory inputs detected in the amygdala (left) and excitatory outputs from the IC to the amygdala (right) in percent of total in- or output, respectively (aIC, red; mIC, green; pIC, blue). Data is shown as average ± SEM. n = 3 mice per condition. b) Input cell density (top row) and percent output density (bottom row) plots along the anterior-posterior axis covering the entire amygdala. We selected aBLA, pBLA, CeA and APir to provide the areas with most differences
2.1.5 IC – Striatum Connectivity
In higher mammals, cortico-striatal projections have been shown to be topographically organized166,167 and play a role in goal-oriented behavior and initiation of movements168. As I
wanted to investigate the regulation of motivated behavior by the IC, the connectivity to the ventral striatum was of particular interest to me. As expected, there were no neurons projecting from striatum to cortex (Figure 3), and I therefore solely describe IC-to-striatum outputs in the
following.
In accordance with a recent study169, ventro-lateral parts of the striatum were more heavily
targeted by IC projections than dorsal parts (Figure 7b). Notably, the vast majority of projections
to the striatum arose from the aIC, which were high in density and spanning the ventro-lateral CPu along its entire rostro-caudal extent. Despite their low relative percentage of total outputs, the IPAC as well as the NAcc were densely innervated by aIC (Figure 7d).
The mIC and pIC also innervated the CPu, but much weaker than the aIC (ca. 5-fold lower). Yet, both mIC and pIC densely projected to the IPAC (to around 60% density). The NAcc was also innervated by the pIC, although much weaker than from mIC or aIC. In the second part of this thesis, I have functionally analyzed the pIC→NAcc projection using optogenetics (see Chapter 2.3) and could show that in spite of the relatively weak innervation, as compared to the mIC or
aIC projection, consummatory behaviors could be potently interrupted.
Overall, comparing mIC and pIC to each other, they displayed a very similar connectivity pattern to the striatum with 9% and 11% of total output, respectively. Remarkably, the aIC projection to the striatum represents the strongest output out of all regions innervated by aIC (31.8%). To summarize, there is a noticeable difference in the projections to the striatum along the rostro-caudal axis of the IC, with the aIC being the most strongly connected part of IC.
Figure 7. IC-striatum connectivity. a) Coronal sections depicting the striatum with its subregions.
b) Representative images from excitatory outputs (eYFP-positive neurons). Note the strong innervation of CPu, NAcc and NAcSh by the aIC. Different Bregma levels are shown for each aIC, mIC and pIC, as indicated on the images. Scale bar = 500 µm. c) Comparison of excitatory outputs from the three IC subdivisions to the striatum in percent of total output (aIC, red; mIC, green; pIC, blue). Values are given as percentage of total pixels. Data shown as average ± SEM, n = 3 mice per condition. d) Plots depict the density of IC innervation along the anterior-posterior axis of the striatum. n = 3 mice per condition, data shown as average ± SEM.
2.2 Functional Characterization of Posterior Insular Cortex
Having mapped the connectivity of the mouse insular cortex, I next wanted to investigate its implications in behavior. As has been suggested previously, the flow of information within the IC starts at the pIC and is progressively transmitted to the aIC40,77. Craig suggested that sensory
information about the body is first mapped in pIC and then subsequently remapped and integrated with emotional information in mIC and aIC77,170. My rabies tracings in part support this view, as
the pIC receives the majority of its inputs from sensory cortex (S1, S2, Pir, Au2), whereas the aIC connectivity is biasing motor related areas. Thus, I chose to functionally investigate the pIC - the potential primary entry point into the insula.
2.2.1 Optogenetic Stimulation of IC Induces a Mixture of Aversive
Behaviors
To characterize the functional implications of the pIC on behavior, I used optogenetics to first broadly stimulate its activity. I infected excitatory pyramidal neurons of the pIC by bilaterally injecting 150 nl of AAV coding for ChR2 under the CamKIIα promotor (AAV2/5-CaMKIIa-hChR2(H134R)-EYFP) and implanted optic fibers 500 µm above the injection sites (Figure 8a).
As my control condition, I used mice carrying bilateral optic fiber implants and that were injected with the same viral vector as for ChR2-mice but expressing only the fluorophore.
I titrated the stimulation parameters and found that 20 Hz stimulation often led to direct and mixed behavioral effects, which became more frequent upon consecutive stimulations (Figure 8b,c). A
typical response pattern began with an interruption of ongoing locomotion, which quickly became freezing. Further stimulation elicited backward escape movements, movements of the jaw or salivation. This would build up with additional stimulations to eventually display very severe behavioral effects, such as jumping, vocalizing, lying on the side in a crouching position. In one case, an animal even died shortly after stimulation.
Figure 8. Bilateral optogenetic stimulation of pIC elicits various stimulation-induced behaviors.
a) Optogenetic virus injections and optic fiber placements above the pIC. AP, anterior-posterior distance from Bregma. GI, granular insular cortex; DI, dysgranular insular cortex; AIP, posterior agranular insular cortex. b) Photostimulation triggered defensive behaviors in ChR2-expressing mice. Rows are reactions of representative individual mice to successive stimulations. Experiments were terminated after 5 minutes or when reaching severely aversive responses (animals 2-6). c) Quantification of light-evoked behavioral responses upon repeated 20 Hz stimulations (n = 15 mice / group derived from two independent experiments with similar results, two-way RM ANOVA, group (opsin) effect, F (1, 28) = 857.6, p < 0.0001; behavior effect, F (6, 168) = 22.63, p < 0.0001, group x behavior interaction, F (6, 168) = 22.69, p < 0.0001; Bonferroni post hoc analysis, grooming *p = 0.0224, jaw movements, stopping, escape and freezing ****p < 0.0001, salivation (p = 0.8581) and paws-to-mouth (p = 0.5988) were not significantly altered). d) Example LFP trace from one ChR2-expressing animal during unilateral photostimulation-induced behaviors. Laser activations (473 nm, 1 s, 20 Hz, 5 ms pulse width, minimum of 4 s ISI) are indicated above the trace (blue) and observed behaviors are shown in color code below. Note the high amplitude LFP signal towards the end of the trace. Example LFP traces from different photostimulation-induced behaviors. Data from Panel d) was created and analyzed by Alexandra Klein. Panels a)-d) are adapted from my publication171 as permitted by the author reuse rights of
Nature Springer.
Especially the last stages of this stereotyped sequence resembled seizure-like activity. Indeed, simultaneous local field potential (LFP) recordings during optogenetic stimulations, which were performed by my colleague Alexandra Klein, detected seizure-like LFP signatures during the jumping and crouching behaviors (Figure 8d).
